Multi-Agent Email & JSON Intelligence System

This is a multi-agent system that accepts email (.eml) files or JSON files, extracts structured and useful information from them using LLMs, and logs everything into a Redis-based memory system.

Features

• Classifier Agent: Routes incoming documents (emails or JSON) to the correct processing agent
• Email Agent: Extracts metadata and summaries from emails
• JSON Agent: Uses an LLM to normalize unstructured JSON into a predefined schema (e.g., invoice format)
• Redis Memory: All results are stored and organized by document and thread in Redis

Installation & Setup

1. Clone the repository

git clone <url>

OR If using VSCode use command pallate (By CTRL + SHIFT + P ) then type git -> select git:clone -> paste the url of repo -> select forlder and ENTER.

2. Install dependencies

Make sure you're in a virtual environment, then run:

pip install -r requirements.txt

3. Set up Redis server

• Recommended: Use WSL (Windows Subsystem for Linux) if on Windows
• Alternatively, you can install Redis using official Redis installation docs
• Run the server and in the config add the port, host and database
• Then add your OPENAI_API_KEY in the config.py

Running the System

1. Test with a JSON file

python main.py sample_inputs/tes_json.json

2. Test with an email file

python main.py sample_inputs/test_email.eml

The system will:

• Print extracted structured output in the terminal
• Log documents and threads in Redis (with doc:<id> and thread:<id> format)

Debug & Monitoring

You can inspect stored documents via:

• RedisInsight GUI (recommended)

Requirements

Check requirements.txt — includes:

• redis — Redis client for Python
• PyMuPDF (fitz) — for PDF/EML parsing
• openai — for LLM-based JSON extraction

Project Structure

project/
├── agents/
│   ├── classifier_agent.py
│   ├── email_agent.py
│   └── json_agent.py
├── memory/
│   └── redis_memory.py
├── utils/
│   ├── file_loader.py
│   └── llm_helper.py
|
├── sample_inputs/
│   ├── tes_json.json
│   └── test_email.eml
├── output_samples/
│   ├── screenshots
|   |   └──(screenshots of output)
|   └── Demo.mp4
├── config.py
├── main.py
├── README.md
└── requirements.txt

Output Samples

The folder output_samples/ contains screenshots,demo video and sample outputs of:

• Terminal logs showing extracted information
• RedisInsight views of stored documents and threads

This folder helps in verifying the system's functionality visually.

Notes

• Make sure Redis server is running before executing main.py
• main.py automatically detects the input file type (email or JSON) and routes accordingly
• The system can be extended with new agents (e.g., PDFAgent) by plugging into the same interface.
• The classifier agent can be extended to take one ,ore input thread_id to add conversation of a particular thread for easy access with some modification and simple logic